Enable AVX2 avg HBD function.

Speed-up of 0.5% to 2.88% for 10-bit encoding as measured on the first
 10 frames of Bosphorus 3840x2160 10bit using --tiles 16 --threads 16
 when compared to commit 8b930d2 on a 32-core 3970x:

            Before                  After               Delta FPS
      Real (s)   User (s)    Real (s)   User (s)    Real (%)   User (%)
 s0   358.144   4041.872     354.193   3991.381        1.12       1.26
 s1    57.760    616.687      57.023    608.007        1.29       1.43
 s2    36.050    345.250      35.594    340.669        1.28       1.34
 s3    31.198    303.964      30.886    300.703        1.01       1.08
 s4    30.988    302.335      30.664    299.023        1.06       1.11
 s5    29.363    279.151      29.051    275.367        1.07       1.37
 s6    17.947    157.552      17.706    154.048        1.36       2.27
 s7    17.384    148.874      17.149    146.637        1.37       1.53
 s8    16.863    148.554      16.676    144.990        1.12       2.46
 s9    14.415    130.140      14.344    129.308        0.50       0.64
s10    10.974     66.348      10.774     64.491        1.86       2.88

Author: negge

src/asm/x86/mc.rs

@@ -475,6 +475,11 @@ extern {
     dst: *mut u8, dst_stride: libc::ptrdiff_t, tmp1: *const i16,
     tmp2: *const i16, w: i32, h: i32,
   );
+
+  fn rav1e_avg_16bpc_avx2(
+    dst: *mut u16, dst_stride: libc::ptrdiff_t, tmp1: *const i16,
+    tmp2: *const i16, w: i32, h: i32, bitdepth_max: i32,
+  );
 }
 
 cpu_function_lookup_table!(
@@ -483,7 +488,11 @@ cpu_function_lookup_table!(
   [(SSSE3, Some(rav1e_avg_ssse3)), (AVX2, Some(rav1e_avg_avx2))]
 );
 
-cpu_function_lookup_table!(AVG_HBD_FNS: [Option<AvgHBDFn>], default: None, []);
+cpu_function_lookup_table!(
+  AVG_HBD_FNS: [Option<AvgHBDFn>],
+  default: None,
+  [(AVX2, Some(rav1e_avg_16bpc_avx2))]
+);
 
 #[cfg(test)]
 mod test {